Chance or coincidence? AMD releases the Athlon 64, specifically targeted at the consumer performance market, and only days later Intel announces a secret processor aimed squarely at the same MHz-hungry community: the Pentium 4 Extreme Edition (P4EE). That's no accident.
Essentially the P4EE is Intel's answer to AMD's new 64-bit processors - and it's not their 64-bitness that Intel's after, it's the high 32-bit performance of the new chips. Intel's solution is also uncharacteristically blatant for the semiconductor supergiant. Instead of offering a new architecture, Intel has essentially grabbed one of its high-end workstation-class Xeon MP processors (Gallatin core) with 2MB of full speed L3 cache, ripped out the multiprocessor capability, and slapped it in the same 478-pin package as the perfunctory P4, and pow! The Pentium 4 Extreme Edition. And that's really all there is to it.
One thing to note, though, is the Pentium 4's caches are inclusive, not exclusive like with the Athlon. This means all the data in the Pentium 4's L2 cache is replicated in the L3 cache - for a grand total of 2MB of cached data.
In terms of performance though, the P4EE's got more than just a few extra ponies under the hood. Given the Pentium 4's long 20 stage pipeline, its bugbear has always been when an instruction gets half way down the pipe only to be discarded because of branch misprediction. If that happens, the instruction is dumped, Neo-style, along with all other operations that were dependent on it. This can cause the entire pipeline to be flushed, and new instructions to be sought from the memory. If the necessary data happens to be in the main system memory, then the CPU has to wait for it to be loaded. If the data is in the cache, it can be accessed far faster, and the CPU can get back down to business sooner. As a result, the Pentium 4 has always seen a significant performance benefit from an increase in cache size, such as the jump to 512KB in the Northwood.
And the P4EE is no exception. Our benchmarking, using SYSmark2002, Photoshop and Quake III: Arena show that the P4EE is now the crowned king of 32-bit applications, and by no small margin. Besides the G5 stonking along in Photoshop, the P4EE leaves the others behind.
So, if you want the fastest, the choice is easy. Well, at least until you get the bill. Pricing is still yet to be confirmed, but it's looking to be around US$925 each in 1,000 quantities, which - using the PC Authority patented Intel-US-pricing-in-volume-conversion-metric of ‘double it and add some' - equates to around $2,000 in the local tender. That makes even the G5 look price competitive. I can't imagine Intel expects to sell bucketloads of P4EEs, but for them it's always nice to know you have the fastest processor out there just in case a wayward Howard Hughes decides to get into gaming.